{"id":108221,"date":"2026-07-17T11:17:20","date_gmt":"2026-07-17T10:17:20","guid":{"rendered":"https:\/\/routenote.com\/blog\/?p=108221"},"modified":"2026-07-17T11:17:22","modified_gmt":"2026-07-17T10:17:22","slug":"suno-hack-reveals-ai-music-training-data-scraped-from-youtube-deezer-and-genius","status":"publish","type":"post","link":"https:\/\/routenote.com\/blog\/suno-hack-reveals-ai-music-training-data-scraped-from-youtube-deezer-and-genius\/","title":{"rendered":"Suno hack reveals AI music training data scraped from YouTube, Deezer and Genius"},"content":{"rendered":"\n

A breach on AI music platform Suno has reportedly revealed that its models were trained using content from major platforms. Here\u2019s what happened and what it means.<\/span><\/h2>\n\n\n\n

Fresh questions are being asked about how AI music companies train their models after a reported hack of Suno revealed details of the datasets behind its AI music generator.<\/p>\n\n\n\n

According to a report<\/a> by 404 Media, source code obtained from a security breach suggests Suno collected music and metadata from platforms including YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound and others to help train its AI models. The findings could add weight to ongoing copyright lawsuits already facing the company.<\/p>\n\n\n\n

The revelations come as the debate around AI-generated music, copyright and creator rights continues to intensify.<\/p>\n\n\n\n

What was revealed?<\/h3>\n\n\n\n

According to 404 Media, a hacker accessed Suno’s internal source code along with customer information and shared portions of that data with the publication. The leaked code reportedly references several sources used when building Suno’s training datasets, including: YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound, International Music Score Library Project (IMSLP), and Podcast RSS feeds.<\/p>\n\n\n\n

The code also appears to log how much content was collected from each source. According to the report, Suno’s datasets included more than 113,000 hours of YouTube Music content, 152,000 hours of tagged YouTube audio, over 17,000 hours from Genius and more than 12,000 hours from Deezer.<\/p>\n\n\n\n

404 Media also reports that some internal code referenced searching for a cappella versions of songs on YouTube, potentially to isolate vocal performances during training.<\/p>\n\n\n\n

While much of this aligns with previous allegations made against Suno, the leaked material offers one of the clearest public glimpses yet into how the company’s datasets may have been assembled.<\/p>\n\n\n\n

Suno says it trained on publicly available music<\/h3>\n\n\n\n

The findings aren’t entirely at odds with Suno’s previous public statements. Suno has already acknowledged<\/a> in court filings that its AI models were trained using music available across the open internet.<\/p>\n\n\n\n

\n
Suno admits it used copyrighted music (blames majors)<\/a><\/blockquote>